Date: Mon, 5 May 2003 15:08:05 +0200 From: Alexander Leidinger <Alexander@Leidinger.net> To: Wesley Morgan <morganw@chemikals.org> Cc: cvs-all@freebsd.org Subject: Re: cvs commit: src/lib/libpthreadMakefilesrc/lib/libpthread/thread thr_attr_get_np.c thr_attr_init.c thr_attr_setcreatesuspend_np.c Message-ID: <20030505150805.6f56bb1c.Alexander@Leidinger.net> In-Reply-To: <20030418113144.K93085@volatile.chemikals.org> References: <Pine.GSO.4.10.10304180221230.4419-100000@pcnet1.pcnet.com> <20030418113144.K93085@volatile.chemikals.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 18 Apr 2003 11:34:59 -0400 (EDT)
Wesley Morgan <morganw@chemikals.org> wrote:
> Ok, I'm a but confuzzled by all these threading libraries. Now we have
> libc_r, libthr, and libpthread? How can I alter the behavior of "-pthread"
Yes.
> to use a different library? As far as I can tell, only libc_r is actually
There's no official way to alter the behavior at compile time. Either
you replace libc_r by e.g. libthr, or you link against one of it by
hand. (... or you modify the binary by hand (s/libc_r/libthr/))
> linked in with a buildworld, so when is it planned to make these
> available? What are the advantages and disadvantages? Which library is
> more complete? Can someone do a summary of some sort to help out the
> consumers?
All of those libs are supposed to implement the POSIX Thread API.
libc_r works completely in userland, so you don't get a benefit from
more than one CPU in _one_ program (e.g. a program with 2 independent
threads doesn't use 2 CPUs at the same time, even if it could).
libthr and libpthread are based upon the KSE framework in the kernel.
libthr uses a 1:1 model, every POSIX thread is mapped to a kernel
thread, the threads in the example program above could run on 2 CPUs at
the same time. libpthread uses a M:N model (not as easy to implement as
a 1:1 model), think of it as of a combination of libthr and libc_r. Not
every POSIX thread is a kernel thread, M POSIX threads are mapped into N
kernel threads. The rationale behind this is, that there are a lot of
programs with threads which aren't independent, so you don't need to
have a kernel thread for every POSIX thread (crossing the barrier
between the userland an the kernel is expensive).
libc_r is widely tested but doesn't deliver the max. performance, libthr
and libpthread aren't widely tested but are supposed to perform better
than libc_r on SMP machines.
Bye,
Alexander.
--
If Bill Gates had a dime for every time a Windows box crashed...
...Oh, wait a minute, he already does.
http://www.Leidinger.net Alexander @ Leidinger.net
GPG fingerprint = C518 BC70 E67F 143F BE91 3365 79E2 9C60 B006 3FE7
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030505150805.6f56bb1c.Alexander>